[analyzer][tests] Fix issue comparison script

[vsavchenko]

Summary

When newer build has duplicate issues the script tried to
remove it from the list more than once. The new approach
changes the way we filter out matching issues.

[NoQ]

Thx!~

clang/utils/analyzer/CmpRuns.py

401–402

If old and new were sets from the beginning we could have turned this into old = old - common etc. and it would have been kinda neat.

clang/utils/analyzer/CmpRuns.py

 import re
 import sys
 
 from math import log
 from collections import defaultdict
 from copy import copy
 from enum import Enum
 from typing import (Any, DefaultDict, Dict, List, NamedTuple, Optional,
-                    Sequence, TextIO, TypeVar, Tuple, Union)
+                    Sequence, Set, TextIO, TypeVar, Tuple, Union)
 
 
 Number = Union[int, float]
 Stats = Dict[str, Dict[str, Number]]
 Plist = Dict[str, Any]
 JSON = Dict[str, Any]
 # Diff in a form: field -> (before, after)
 JSONDiff = Dict[str, Tuple[str, str]]
def compare_results(results_old: AnalysisRun, results_new: AnalysisRun,
     common_locations = locations_old & locations_new
 
     for location in common_locations:
         old = diags_old[location]
         new = diags_new[location]
 
         # Quadratic algorithms in this part are fine because 'old' and 'new'
         # are most commonly of size 1.
-        for a in copy(old):
-            for b in copy(new):
+        common: Set[AnalysisDiagnostic] = set()
+        for a in old:
+            for b in new:
                 if a.get_issue_identifier() == b.get_issue_identifier():
                     a_path_len = a.get_path_length()
                     b_path_len = b.get_path_length()
 
                     if a_path_len != b_path_len:
 
                         if histogram == HistogramType.RELATIVE:
                             path_difference_data.append(
                                 float(a_path_len) / b_path_len)
 
                         elif histogram == HistogramType.LOG_RELATIVE:
                             path_difference_data.append(
                                 log(float(a_path_len) / b_path_len))
 
                         elif histogram == HistogramType.ABSOLUTE:
                             path_difference_data.append(
                                 a_path_len - b_path_len)
 
-                    res.add_common(a)
-                    old.remove(a)
-                    new.remove(b)
+                    res.add_common(b)
+                    common.add(a)
 
-        for a in copy(old):
-            for b in copy(new):
+        old = filter_issues(old, common)
+        new = filter_issues(new, common)
+        common = set()
+
+        for a in old:
+            for b in new:
                 if a.is_similar_to(b):
                     res.add_changed(a, b)
-                    old.remove(a)
-                    new.remove(b)
+                    common.add(a)
+                    common.add(b)
+
+        old = filter_issues(old, common)
+        new = filter_issues(new, common)
 
         # Whatever is left in 'old' doesn't have a corresponding diagnostic
         # in 'new', so we need to mark it as 'removed'.
         for a in old:
             res.add_removed(a)
 
         # Whatever is left in 'new' doesn't have a corresponding diagnostic
         # in 'old', so we need to mark it as 'added'.
def compare_results(results_old: AnalysisRun, results_new: AnalysisRun,
     if histogram:
         from matplotlib import pyplot
         pyplot.hist(path_difference_data, bins=100)
         pyplot.show()
 
     return res
 
 
+def filter_issues(origin: List[AnalysisDiagnostic],
+                  to_remove: Set[AnalysisDiagnostic]) \
+                  -> List[AnalysisDiagnostic]:
+    return [diag for diag in origin if diag not in to_remove]
+
+
 def compute_percentile(values: Sequence[T], percentile: float) -> T:
     """
     Return computed percentile.
     """
     return sorted(values)[int(round(percentile * len(values) + 0.5)) - 1]
 
 
 def derive_stats(results: AnalysisRun) -> Stats: